Visualization of Semantic Networks

This is the code for visualizing networks using some out-of-the box techniques that are available through networkx and using a popular manifold discovery algorithm UMAP.

Simple Visualizations

We start out here just by visulazing with the defaults for all AD patients.

Now lets look at the normative control graphs

A bit of interpretation

I would say these graphs look pretty good if they low number of nodes.

However they still don't give insights into how alzhiemers patients produce lists.

UMAP

Working with UMAP for a simple example.

Exploration of Adjacency Matrices

Here we explore the usage of adjacency matrices as the input into UMAP. As one can see it doesn't perform to well. Likely due the the high dimensionality of considering all possible nodes as input.

Attempting UMAP

Here I am attempting do do UMAP for a hyperparameter sweep of the number of neighbors and of the minimum distance matrix. Since the minimum distance is greater than the maximum possible spread at points the code fails. However it continues to the next iteration.

the total number of responses, perseveration rate, mean degree, , and diameter. Four factors were individually significant () while one ( diameter) was not ().

Trying all visualizations beyond the default layout in Networkx

Here I'm trying it for the first six NC patients.